Smoking Sales in the United States
Jay-Ho Chung
11/4/2019
Smoke and Mirrors: What Drives the Sales of Cigarettes?
The question we are examining within this article is the sales of cigarettes per capita.
The history of smoking cigarettes dates back to 5000 BCE and became very widespread at the turn of the 17th century with the growth of tobacco within the Amerian colonies, specifically Virginia. This spurred the growth of cigarettes within Europe as various individuals proclaimed the health benefits of tobacco. Even then, however, some people had doubts. One such person was King James I who noted its dangers to the brain and the lungs. Yet, smoking’s harmful effects took another four hundred years to prove and many people continued to smoke, from pipes to cigarettes.
In the mid-20th century, scientists began to find harmful effects from smoking which were published heavily within the United States towards the late 20th century. But what happened to the sales of cigarettes in the late 20th century? What exactly drives the sales of cigarettes? It is difficult to measure the impact of legislation in a variable, but the effect of different factors such as price and disposable income will be investigated. The question comes into as to what exactly drives the sales of cigarettes. Is it the price of cigarettes or disposable income? We will examine what factors drive the sales of cigarettes and analyze both the strengths and weaknesses of the model.
library(tidyverse)
library(ggthemes)
library(gridExtra)
library(xtable)Analysis
Data, Broken Down
This data set, smo measured cigarette consumption from 1963 to 1992 in 46 states. The data set is available here. There were several categorical variables within this data set.
state assigned a unique number to each state.
year marks the year, starting from 1963 (denoted as 63) to 1992 (denoted as 92).
price describes the price of a pack of cigarettes. r_pri adjusted the prices for real prices. It is important to adjust for real prices because it gives a better understanding of how to compare the prices from year to year as it removes inflation as a possibility in the increase in sales. Basically, if inflation were not accounted for, then inflation could have been an unaccounted factor behind the increase in price of cigarettes. pop is the population within each state in thousands. pop16 illustrates the population above 16.
cpi describes the consumer price index, a measure of inflation with respect to 100 as the base CPI in 1983.
ndi is the disposable capital income per person, but r_ndi (disposable income adjusted for inflation) was used. sales provides the number of sales of cigarettes in pack per person in each state.
pimin demonstrates the “minimum price in adjoining states per pack of cigarettes.” r_pimin was used to adjust for inflation.
Unfortunately, the state names were not provided in this dataset. Therefore, 1970 Census Data was compared from was used to identify each of the states. This will be denoted as sta_name.
So, within this dataset, the state number follows as such - 1: Alabama, 2: (No Alaska), 3: Arizona, 4: Arkansas, 5: California, 6: (No Colorado), 7: Connecticut, 8: Delaware, 9: D.C., 10: FL, 11: GA, 12: HI (No Hawaii), 13: Idaho, 14: Illinois, 15: Indiana, 16: Iowa, 17: Kansas, 18: Kentucky, 19: Louisiana, 20: Maine, 21: Maryland, 22: Massachusetts, 23: Michigan, 24: Minnesota, 25: Mississippi, 26: Missouri, 27: Montana, 28: Nebraska, 29: Nevada, 30: New Hampshire, 31: New Jersey, 32: New Mexico, 33: New York, 34: (No North Carolina), 35: North Dakota, 36: Ohio, 37: Oklahoma, 38: Oregon, 39: Pennsylvania, 40: Rhode Island, 41: South Carolina, 42: South Dakota, 43: Tennessee, 44: Texas, 45: Utah, 46: Vermont, 47: Virginia, 48: Washington, 49: West Virginia, 50: Wisconsin, 51: Wyoming.
state_name <- data.frame(state = c(1, 3, 4, 5, 7, 8, 9, 10, 11, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45,46,47,48,49,50,51), sta_name = c("Alabama", "Arizona", "Arkansas", "California", "Connecticut", "Delaware", "D.C.", "Florida", "Georgia", "Idaho", "Illinois", "Indiana", "Iowa", "Kansas", "Kentucky", "Louisiana", "Maine", "Maryland", "Massachusetts", "Michigan", "Minnesota", "Mississippi", "Missouri", "Montana", "Nebraska", "Nevada", "New Hampshire", "New Jersey", "New Mexico", "New York",
"North Dakota", "Ohio", "Oklahoma", "Oregon", "Pennsylvania", "Rhode Island", "South Carolina", "South Dakota", "Tennessee", "Texas", "Utah", "Vermont", "Virginia", "Washington", "West Virginia", "Wisconsin", "Wyoming"))
#state_name #Some join will be required
smo_mod <- smo %>%
left_join(state_name, by = "state") #Left join to add the names to smoVisualizing the Data
So what exactly impacts the sales of cigarettes? The first hypothesis was that it would be just real price and real disposable income.
We first graphed both the real price and disposable income to see how that looked. It looked quite linear, so we then checked the sales of cigarettes.
It also appears as though there is a positive relationship between real disposable income per capita and years: as time goes on, the disposable income for each person has increased.
With this being said, there was not a linear pattern within the sales of cigarettes per capita (below), and it is unclear if disposable income and prices are the only factors behind the real sales of cigarettes. A multiple regression was therefore run as it appeared as though there were more factors beyond just price and disposable income.
smo_mki %>%
ggplot(aes(x= year, y = sales)) +
#theme_economist() +
geom_point(aes(group = factor(state), color = (sta_name)), size = 6) +
labs(x = "Years" , y = "Sales of Cigarettes Per Capita",
#X axis code does not work
title = paste0("Sales of Cigarettes Per Capita v. Years"), subtitle = "United States, 1963-1992") +
theme(plot.title = element_text(size = 50) ,
axis.text.x = element_text(size = 50) ,
axis.text.y = element_text(size = 50),
axis.title.x = element_text(size = 50) ,
axis.title.y = element_text(size = 50),
legend.text = element_text(size = 20),
plot.subtitle = element_text(size = 40)) 
Note that within the sales plot 70 indicates 1970, 80 indicates 1980, and so on.
Two interesting things did show up within this sales plot. Utah seems to have lowest amount of cigarette sales, which could possibly be attributed to the large Mormon population within the state. New Hampshire has a high (and fluctuating) value in terms of sales of cigarettes. Yet, overall, it does not appear as though the sales of cigarettes is solely based off of the disposable income of the people in the state and the real prices of cigarettes.
A table of the sales of cigarettes per capita for every state in every year was included below. This is a quite a long table, but it illustrates what the plot showed above.
xtable(smo_sales)#xtable(smo_sales2)National Model (Multiple Regression Model 1)
It appears as though the first hypothesis failed, so the second was that the sales of cigarettes would be dependent on many varying factors. This includes the real prices, real disposable income, state name, year, population over 16, and real price of cigarettes of adjoining states. This model was looking at the United States as a whole, so it is the national model.
So what we are running here is called a linear regression. This is basically just a tool that statisticians use to see if a variable (in this case the sales of cigarettes) can be explained be other variables (real prices, real disposable income, state name, year, population over 16, etc.)
xtable(step_fit)#plot(fit)The assumptions of linearity and the nearly normal condition, however, needed to be checked. The residuals plot illustrates that the model has some difficulty or a large amount of error as the prices increased. In other words, this model has some faults.
With all this stats jargon put aside, I’m basically just saying that I needed to test if the model ws linear, and this thing called a residuals plot shows if the model is actually linear – if it a linear model would actually make sense to use.
The model had an adjusted \(R^2\)-squared value of 0.8081 This simply showed that 80.81 percent of the variability in the sales of cigarettes can be explained by real prices, real disposable income, state name, year, population over 16, and real price of cigarettes of adjoining states.
All you need to know from that is that this statistical model is fairly strong and appears as though it is appropriate for analysis.
\[ \hat{Sales of Cigarettes} = -1.06 * Real Price + 0.00136 * Real Disposable Capital Income - 0.430 * Year + 0.229 * Minimum Price in Adjoining States \]
\[ -0.425 * AZ + 7.675* AK + 0.953 * CA + 12.17 *CT + 45.85 * D.C. + 39.92 * DE + 23.06 * FL \] \[ + 7.56 * GA + 16.05 * IL + 22.54 * IN + 2.60 * IA -1.53 * KS + 52.60 * KY + 14.22 * LA \]
\[ + 25.12 * ME + 6.29 * MD + 14.46 * MA + 18.39*MI + 16.18 * MO + 64.26 * NV + 113.84 * NH + 12.73 * \]
\[ NJ -16.45 * NM + 12.47 * NY + 13.00 * OH + 10.90 * OK + 6.23* PA + 23.06 * RI - 2.82 * SC + \]
\[ 10.00 * TN + 6.45 * TX - 42.96 * UT + 29.21 * VT + 8.56 * VA + 8.69 * WV + 0.0938 WI+209.17 \]
The model above describes how the sales of cigarettes is impacted. So, basically, the real price of cigarettes decreases the sales of cigarettes by 1.06, the real disposable income increases the sales and such forth.
It’s important to note that the states varied in their impact upon the sales of cigarettes, as some states were less important than others on the national level.
California, Connecticut, D.C., Delaware, Florida, Georgia, Illinois, Indiana, Kentucky, Louisiana, Maine, Maryland, Massachusetts, Michigan, Missouri, Montana, Nevada, New Hampshire, New Jersey, New Mexico, New York, North Dakota, Ohio, Oklahoma, Pennsylvania, Rhode Island, Tennessee, Utah, Vermont, Virginia, Wisconsin, and Wyoming appear to be the most significant contributors in terms of sales of cigarettes.
Iowa, Kansas, and Texas appeared to be were less significant than those states (significant in the 99.9 percent). Arizona, Arkansas, and West Virginia were also significant in the the 95 percent, so at a lesser degree than Iowa, Kansas, and Texas.
The other states did not appear to have an effect upon the sales of cigarettes per pack (Idaho, Minnesota, Mississippi, Montana, Nebraska, North Dakota, South Dakota, Washington and Wisconsin).
Factors that Decrease Sales
Real price (by 1.06) Year (by 0.430) Arizona (by 0.425) Kansas (1.53) New Mexico (16.45) South Carolina (2.82) Utah (42.96)
Factors that Increase Sales
Real disposable income (0.00136) Minimum Price in Adjoining States (0.229) California, Connecticut, D.C., Delaware, Florida, Georgia, Illinois, Indiana, Kentucky, Louisiana, Maine, Maryland, Massachusetts, Michigan, Missouri, Montana, Nevada, New Hampshire, New Jersey, New Mexico, New York, North Dakota, Ohio, Oklahoma, Pennsylvania, Rhode Island, Tennessee, Utah, Vermont, Virginia, Wisconsin, Wyoming, Iowa, Texas, and West Virginia all increased the sales of cigarettes by each of their coefficients (summarized in the table and in the model).
Another model was also created to break it down into decades, but the residuals plot looked roughly similar (each model had difficulty predicting as the sales increased). Therefore, this decade model was not included
Model Strengths and Weaknesses
This model has some strengths and weaknesses; it appears to be able to predict the sales fairly accurately and takes into account many variables. This model seems to have issues predicting when sales are very high within states. This is likely because of human behavior and this model simply cannot account for this. One possible aspect of human behavior is the fact that people follow social norms; if different states or communities are more accepting of smoking, then more people may be more likely to smoke and this cannot be described by any of the variables within the data.
This model is also unable to account for legislation and public efforts to discourage smoking. In 1984, Congress forced cigarette companies to attach labels on the health risks of cigarettes, which may have also affected the sales. This is unable to be accounted for within the model.
Regional Model (Multiple Regression Model 2)
How can we begin to improve this model based off all these issues with the model above? Well, maybe splitting the data into regions. By splitting it into regions, we can then – maybe – better capture an idea of what drives smoking within the United States.
A question, however, comes up as to whether splitting this into regions and trying to predict the sales will give a more accurate idea of the sales. This was based on the thinking that regions would follow smoking patterns more than the entire nation.
A model for each region was then created: the idea being that the sales would be dependent on the real price of a pack of cigarettes, the real income per capita, the year, the population over 16, the real minimum price of the surrounding areas, and the state. A model was made for each region, New England, Mid-Atlantic, Southeast, Midwest, and the West.
reg_plot <- function(reg_id) {
smo_reg %>%
filter(region == reg_id) -> reg_dat
ggplot(reg_dat, (aes(x = year, y = sales))) +
geom_point(aes(group = factor(state), color = (sta_name)), size = 6) +
labs(title = paste0("Sales of Cigarettes Per Pack Per Capita in ", reg_id ," by State v. Years")) +
#theme_economist() +
theme(plot.title = element_text(size = 50) ,
axis.text.x = element_text(size= 50) ,
axis.text.y = element_text(size = 50),
axis.title.x = element_text(size = 50),
axis.title.y = element_text(size = 50),
legend.text = element_text(size = 25),
plot.subtitle = element_text(size = 30))
}New England
\[ \hat{Sales of Cigarettes} = -1.34 * Real Price - 0.005 * Real Disposable Capital Income + \] \[ 0.289 * Year - 0.016 * Population Over 16 +0.742 Minimum Price in Adjoining States \] \[ -36.40 *ME+ 28.39 *MA+ 57.81*NH-40.04* RI - 36.81 * VT + 269.81 \]
reg_plot("NE")
This is the resulting model of sales for New England, with the base case being Connecticut. It illustrates that the sales of cigarettes (in packs) per capita in New England decreases by 1.34 as the price of cigarettes increases by one dollar. As disposable income increases by 1000 dollars, the number of packs per capita sold goes down by 5. As the year increases from 1963 to 1992, the number of packs sold per capita increased by 0.289 as the year increased (from 63 to 63). As the legal age populatoin increased, however, the number of sales of cigarettes sold decreased by 16 packs for every thousand people.
Essentially, this is indicating that states with a higher population in New England were less likely to consume less cigarettes. As the minimum price of cigarettes in surrounding areas increased by a dollar, the number of packs sold in New England increased by 0.742.
Maine, relative to Connecticut, sold 36.40 less packs of cigarettes. Rhode Island, 40.04 less, and Vermont with 36.81 less per capita. Massachusetts sold 28.39 more and New Hampshire 57.81 sold more.
To summarize without all the numbers, the factors that increased the sales of cigarettes in New England are the real price in adjoining states, disposable income, year, Massachusetts, and New Hampshire.
The factors that decreased the sales of cigarettes are real price of a pack of cigarettes, real disposable income, population over 16, Maine, Rhode Island, and Vermont.
There are a few things that seem strange about this model, however. It seems strange that the minimum price in other surrounding states brings the sales of cigarettes up, but an increase in the real price of the sales of cigarettes decreases the sales. This could indicate that people going to other states to buy cigarettes, perhaps because of different laws on the legal age for the purchase of cigarettes at the time. In addition, it could mean that the real price does not play a large role in the consideration in the purchase of sales of cigarettes within New England.
Furthermore, perhaps the sales metric is not the greatest measure. It is a measure of sales of packs of cigarettes per capita. This means that it does not accurate how many smokers there are in a state. Suppose that Massachusetts has a higher population than Maine.
Midwest
\[ \hat{Sales of Cigarettes} = -0.90 * Real Price + 0.004 * Real Disposable Capital Income \]
\[ - 0.35 *Year-0.007 *Population Over 16+ 0.13*Real Minimum Price in Adjoining States+ \]
\[ 71.51 *IL+ 53.39 *IN+19.19*IA+11.78*KS+ \]
\[ 65.35 *MI+ 24.85*MN+ 43.89 *MS- 0.289 *ND+ 70.29 *OH- 0.54*SD+ \] \[ 25.90 *WS+ 166.52 \]
reg_plot("MW")
This is the resulting model of sales for the Midwest, with the base case being Idaho.
It illustrates that the sales of cigarettes (in packs) per capita in the Midwest decreases by 0.90 for every dollar increase in the price of cigarettes. As disposable income increases by 1000 dollars, the number of packs per capita sold goes up by 4 packs. As the year increases from 1963 to 1992, the number of packs sold per capita decreased by 0.35.
As the legal age population increased, the number of sales of cigarettes sold decreased by 7 packs for every thousand people. As the minimum price of cigarettes in adjoining states increased by a dollar, the number of packs sold in the Midwest increased by 0.13.
Every state except for North and South Dakota (which sold 0.289 and 0.54 less packs of cigarettes relative to Idaho) sold more packs of cigarettes relative to Idaho. Illinois sold 71.51, Indiana 53.39, Iowa 19.19, Kansas 11.78, Michigan 65.35, Minnesota 24.85, Missouri 43.89, Ohio 70.29, and Wisconsin 25.90 more packs of cigarettes than Idaho.
Southeast
sal_reg("SE")\[ \hat{Sales of Cigarettes} = -0.88 * Real Price + 0.001 * Real Disposable Capital Income \]
\[ - 0.34 *Year-0.006 *Population Over 16 \]
\[ -0.13 *Real Minimum Price in Adjoining States+ 1.04*AK+ \]
\[ 31.82*FL+3.89 *GA+ 47.56*KY+ 12.43 *LA + 2.00 *MS- 30.77 *NS \] \[ -0.71 *OK-3.46*SC+2.82*TN+35.07*TX + 167.50 \]
reg_plot("SE")
This is the resulting model of sales for the Southeast, with the base case being Alabama.
It illustrates that the sales of cigarettes (in packs) per capita in the Southeast decreases by 0.88 sa the real price goes down. As disposable income increases by 1000 dollars, the number of packs per capita sold goes up by 1 packs. As the year increases from 1963 to 1992, the number of packs sold per capita decreased by 0.34. As the legal age population increased the number of sales of cigarettes sold decreased by 6 packs for every thousand people.
As the minimum price of cigarettes in adjoining states increased by a dollar, the number of packs sold in the Southeast decreased by 0.13.
Arkansas sold 1.04, Florida 31.82, Georgia 3.89, Kentucky 47.56, Louisiana 12.43, Missouri 2.00, South Tennessee 2.82, and Texas 35.07 sold more packs of cigarettes than Alabama according to the model.
Oklahoma and South Carolina sold 0.71 and 3.46 less packs of cigarettes than Alabama.
Midatlantic
sal_reg("MA")\[ \hat{Sales of Cigarettes} = -1.15 * Real Price - 0.005 * Real Disposable Capital Income \]
\[ -1.48 *Year+ 0.034 *Population Over 16 + \]
\[ 0.70 *Real Minimum Price in Adjoining States- 23.65 *DE - 137.07 *MD \] \[ -209.90 *NJ- 489.30 *NY- 347.39 *PA- 168.41*VA-96.13 *WV + 371.95 \]
reg_plot("MA")
This is the resulting model of sales for the Midatlantic states, with the base case being Washington D.C.
It illustrates that the sales of cigarettes (in packs) per capita in the Midatlantic decreases by 1.15 for every one dollar increase in real price of a pack of cigarettes. As disposable income increases by 1000 dollars, the number of packs per capita sold goes down by 5.15 packs. As the year increases from 1963 to 1992, the number of packs sold per capita decreased by 1.47. As the legal age population increased, the number of sales of cigarettes sold icnreased by 34 packs for every thousand people.
As the minimum price of cigarettes in adjoining states increased by a dollar, the number of packs sold in the Midatlantic increased by 0.70.
All states sold less cigarettes than D.C. Delaware sold 23.65, Maryland 137.07, New Jersey 209.90, New York 489.30, Pennsylvania 347.39, Virginia 168.41, and West Virginia 96.13 less packs of cigarettes than D.C. according to the model.
Perhaps D.C. was not the base case, as it appears as though D.C. sold the most number of cigarettes out of any of these states.
Northwest
sal_reg("NW")\[ \hat{Sales of Cigarettes} = -0.20 * Real Price + 0.0098 * Real Disposable Capital Income \]
\[ - 1.57 *Year - 0.003 *Population Over 16 \] \[ -0.49 *Real Minimum Price in Adjoining States - 32.44 *MN - 52.97 *WA - 16.69 *WY + 239.25 \]
reg_plot("NW")
This is the resulting model of sales for the Northwest states, with the base case being California.
It illustrates that the sales of cigarettes (in packs) per capita in the Northwest decreases by 0.20 for every one dollar increase in real price of a pack of cigarettes.
As disposable income increases by 1000 dollars, the number of packs per capita sold goes up by 9.8 packs. As the year increases from 1963 to 1992, the number of packs sold per capita decreased by 1.57.
As the legal age of the population increased the number of sales of cigarettes sold decreased by 3.2 packs for every thousand people. As the minimum price of cigarettes in adjoining states increased by a dollar, the number of packs sold in the Northwest decreased by 0.49.
Relative to California, Minnesota sold 32.44, Washington 52.97, and Wyoming 16.69 less cigarettes than California.
There is a problem with this model, however. Just as all the other regions, the sales variable does not necessarily account for the number of smokers. Furthermore, this dataset is missing Colorado – a big consumer of cigarettes. Furthermore, the fact that only four states are in the Northwest means that this may not necessarily be the most useful model.
Southwest
sal_reg("SW")\[ \hat{Sales of Cigarettes} = -0.36 * Real Price + 0.0077 * Real Disposable Capital Income \]
\[ -1.48 *Year - 0.0056 * Population Over 16 \]
\[ -0.33 *Real Minimum Price in Adjoining States + 37.30 * NV - 17.09 *NM \]
\[ -44.80 *UT + 226.60 \]
reg_plot("SW")
Above is the resulting model of sales for the Southwest states, with the base case being Arizona.
It illustrates that the sales of cigarettes (in packs) per capita in the Northwest decreases by 0.36 for every one dollar increase in real price of a pack of cigarettes. As disposable income increases by 1000 dollars, the number of packs per capita sold goes up by 7.7 packs. As the year increases from 1963 to 1992, the number of packs sold per capita decreases by 1.48. As the legal age of the population increased the number of sales of cigarettes sold increased by 5.6 packs for every thousand people. As the minimum price of cigarettes in adjoining states increased by a dollar, the number of packs sold in the Southwest decreased by 0.33
Relative to Arizona, Nevada sold 37.30 more packs of cigarettes per capita. New Mexico and Utah sold 17.09 and 44.80 less packs of cigarettes.
Strengths and Issues with the Regional Models
We should revisit some overall possible issues with this model. There were basically some assumptions made, and they may or not may be correct.
A problem overlapping all of these regional models is that they may be too specific in a certain sense. By specifying the regions so much, there is a certain loss in the generality of the models. We are unable to talk about the United States as a whole as the first multiple regression model.
Another issue is that some states are missing. Colorado, specifically is missing and could be a significant factor in the Northwest and this cannot necessarily be a good measure of the regions as a whole.
With all of this being said, however, the breakdown of each model into the region does make more sense in a few ways. First of all, given factors such as climate or inclination toward smoking could be better accounted for by splitting this up into regions.
Insignificant states within the large multiple regression model become more significant by splitting it into the regional models.
With this being said, there may be an issue. Some of these regions may not be split up in the right way. Is it right to put D.C. in Midatlantic? Is it right to put Texas in the Southeast? These kinds of questions will impact the regional models.
Furthermore, perhaps the sales metric is not the greatest measure. It is a measure of sales of packs of cigarettes per capita. This means that it does not necessarily accurately predict how many smokers there are in a state. Suppose that Massachusetts has a higher population than Maine. Even if the number of smokers in Maine and Massachusetts are the same, the sales of packs of cigarettes per capita would be lower in Massachusetts than Maine. So, therefore, this variable does not necessarily capture the number of smokers accurately.
Conclusion
So what exactly drives the sales of cigarettes? The initial hypothesis was that this would be impacted by real income and real disposable income. Yet, upon further analysis, that was not the case. It was dependent on much more: the state, the minimum price surrounding the area, the year, and the state (multiple regression model 1). Yet, a question came up as the model had some issues as the sales were high. In order to account for this, the data and states were split up into regions to get a better idea of each region.
There are absolutely issues with these two models and they certainly and likely complement each other. Model 1 (National Model) shows a better idea of what drives the sales of cigarettes nationally while Model 2 (Regional Model) demonstrates how the sales of cigarettes can be predicted per region which helped account for some issues within Model 1’s prediction issues as the sales of cigarettes increased.
Some things that would be conducted, if given more time, would be the analysis of the history of smoking legislation of smoking. Perhaps a better model would account for the effect of major legislation of smoking or better graphs would include points at which legislation was passed.
As mentioned in earlier sections, the sales variable itself does not necessarily account for the number of smokers, rather it is a proportion. I would, in the future, try to find a data set that provides the number of cigarettes sold per pack rather than the sales per pack per capita.
Although a lot of statistical information was not included, it has been included in another version so please refer to that for further analysis upon the models and their appropriateness.
References
“Cigarette Consumption Dataset”, R, vincentarelbundock.github.io/Rdatasets/doc/Ecdat/Cigar.html (Web). Accessed November 4, 2019.
Baltagi, B.H. and D. Levin (1992), “Cigarette taxation: raising revenues and reducing consumption”, Structural Changes and Economic Dynamics, 3, p. 321–335.
Baltagi, B.H., J.M. Griffin and W. Xiong (2000), “To pool or not to pool: homogeneous versus heterogeneous estimators applied to cigarette demand”, Review of Economics and Statistics, 82, p. 117–126.
United States Census Bureau,“1970 Census of Population, Advance Reports: General Population Characteristics”, Census Publications, February 1971, https://www.census.gov/library/publications/1971/dec/pc-v2.html (Web), Accessed November 4, 2019.
Yale University Library, “The Rise of Anti-smoking Movements”, Yale University Library Exhibits, http://exhibits.library.yale.edu/exhibits/show/sellingsmoke/antismoking (Web). Accessed November 4, 2019.
Glass, A, “Congress bans airing cigarette ads, April 1, 1970”, Politico Magazine, April 1, 2018, https://www.politico.com/story/2018/04/01/congress-bans-airing-cigarette-ads-april-1-1970-489882 (Web). Accessed November 4, 2019.